In [1]:
%matplotlib inline
import os
import IPython.display as ipd
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import sklearn as skl
import sklearn.utils, sklearn.preprocessing, sklearn.decomposition, sklearn.svm
import utils
sns.set(color_codes=True)
sns.set_context("notebook", font_scale=1)
plt.rcParams['figure.figsize'] = (17, 5)
In [3]:
tracks = pd.read_pickle("clean_data/track.pkl")
tracks.sample(3)
Out[3]:
In [4]:
tracks.dtypes
Out[4]:
In [5]:
tracks.columns
Out[5]:
In [6]:
tracks.shape
Out[6]:
In [7]:
dup_rows = tracks[tracks.duplicated()]
dup_rows.shape
Out[7]:
In [8]:
tracks.isnull().sum()
Out[8]:
MATRIZ DE CORRELACIÓN¶
In [9]:
plt.figure(figsize=(13,10))
sns.heatmap(tracks.corr(),annot=True)
plt.title("Correlation Heatmap")
plt.show()
In [10]:
tracks.genre_top.value_counts()
Out[10]:
Histograma de generos¶
In [11]:
tracks.genre_top.value_counts().nlargest(40).plot(kind='bar', figsize=(10,5))
plt.title("Number of tracks by Genre")
plt.ylabel("Number of tracks")
plt.xlabel("Genre");
Creacion Dummies de generos¶
In [12]:
Data1= pd.get_dummies(tracks.genre_top)
tracks = pd.concat([tracks, Data1], 1)
tracks.shape
Out[12]:
In [13]:
tracks.columns
Out[13]:
In [14]:
sns.barplot(tracks.genre_top.value_counts().index, tracks.genre_top.value_counts(), order=list(tracks.genre_top.value_counts().index))
plt.title("Cantidad de Tracks segun Generos ")
plt.xlabel("Generos")
plt.ylabel("Tracks")
Out[14]:
Correlacion entre variables
In [15]:
#Correlacion entre Variables de echones
lista_echones=['acousticness', 'danceability','energy', 'instrumentalness', 'liveness', 'speechiness', 'tempo','valence']
corr = tracks[lista_echones].corr()
plt.figure(figsize=(13,10))
sns.heatmap(corr,annot=True)
plt.title("Correlation Heatmap Echonest")
plt.show()
corr
Out[15]:
In [17]:
#Correlacion entre Generos
lista_generos=['Blues', 'Classical', 'Electronic', 'Experimental', 'Folk',
'Hip-Hop', 'Instrumental', 'International', 'Jazz',
'Old-Time / Historic', 'Pop', 'Rock']
corr = tracks[lista_generos].corr()
plt.figure(figsize=(13,10))
sns.heatmap(corr,annot=True, cmap='Blues')
plt.title("Correlation Heatmap")
plt.show()
corr
Out[17]:
In [23]:
#Funcion de heatmap por genero
def plot_genero(df, genero):
plt.figure(figsize=(2,5))
sns.heatmap(df.corr()[[genero]],annot=True)
plt.title("Correlation Heatmap Echonest by GENERO: "+genero)
plt.show()
In [24]:
lista_echones=['acousticness', 'danceability','energy', 'instrumentalness', 'liveness', 'speechiness', 'tempo','valence']
for x in lista_generos:
df = tracks[lista_echones +[x]]
plot_genero(df, x)
Visuales de Fechas¶
In [25]:
import time
import datetime
from datetime import date
from bokeh.plotting import figure
from bokeh.io import show, output_notebook
from bokeh.palettes import Category20
from bokeh.models import HoverTool
keys = tracks["genre_top"].unique()
colors = [Category20[20][i] for i in range(20) ]
colormap = dict(zip(keys, colors))
colormap
tracks['date_created'].plot()
plt.figure()
Out[25]:
Visuales de Duracion¶
In [26]:
ax = sns.boxplot(x="genre_top", y="duration", data=tracks, fliersize=2)
plt.title("Duracion segun Genero ")
plt.xlabel("Generos")
plt.ylabel("Duracion")
Out[26]:
In [29]:
def boxplot_echonest (df, echonest):
ax = sns.boxplot(x="genre_top", y=echonest, data= df, fliersize=2)
plt.title("Genero por "+ echonest)
plt.xlabel("Generos")
plt.ylabel(echonest)
plt.show()
In [30]:
for x in lista_echones:
boxplot_echonest (tracks, x)
In [31]:
tracks.genre_top.value_counts().index[0:5]
features = ['acousticness', 'danceability', 'energy', 'instrumentalness', 'liveness', 'speechiness', 'valence', 'genre_top']
full_features = ['acousticness', 'danceability', 'energy', 'instrumentalness', 'liveness', 'speechiness', 'tempo', 'valence']
In [36]:
def compare_two_genres(df, genre_1, genre_2):
_features = ['acousticness', 'danceability', 'energy', 'instrumentalness', 'liveness', 'speechiness', 'tempo', 'valence', 'genre_top']
compare_mask = (df.genre_top == genre_1) | (df.genre_top == genre_2)
_df = df[compare_mask][_features]
sns.pairplot(_df,hue='genre_top', plot_kws = {'alpha': 0.6, 's': 80, 'edgecolor': 'k'});
In [37]:
compare_two_genres(tracks, 'Hip-Hop', 'Pop')
In [38]:
def plot_features_profile(df, genre):
angles = np.linspace(0, 2*np.pi, len(features), endpoint=False)
labels = features
stats = df[df.genre_top == genre][features].mean().values
stats=np.concatenate((stats,[stats[0]]))
fig = plt.figure(figsize=(10,10))
ax = fig.add_subplot(111, polar=True)
ax.plot(angles, stats, 'o-', linewidth=2)
ax.fill(angles, stats, alpha=0.25)
ax.set_thetagrids(angles * 180/np.pi, labels)
ax.set_title( genre + " Profile")
ax.grid(True)
plt.show()
In [39]:
for x in lista_generos:
plot_features_profile(tracks, x)
In [ ]:
# por localidad - ver dist de genero
# por genero - ver localidad
In [42]:
tracks.groupby('genre_top')[features].mean()
Out[42]:
In [50]:
top_localidades=tracks.location.value_counts().head(10)
top_localidades
Out[50]:
In [88]:
lista_top_localidades = top_localidades.index.tolist()
mask_localidades = tracks.location.isin(lista_top_localidades)
tracks[mask_localidades].groupby('location')[lista_generos].sum().plot(kind='bar', figsize=(20,15))
Out[88]:
In [ ]: